FINAL REPORT
to
NATIONAL AERONAUTICS AND SPACE ADMINlSTRA'tlON*Goddard Space Flight CenterGreenbelt, Maryland 20771
CODING FOR RELIABLE SATELLITE COMMUNICATIONS
Grant No.: NAG 5-407
Period: July 1, 1984 to June 30, 1986
Principal Investigators
N.T. Gaarder(January 1 - June 30, 1986)
Shu Lin(July 1, 1984 - December 31, 1985)
Department of Electrical EngineeringUniversity of Hawaii at Manoa
Honolulu, Hawaii 96822
\{\(' (NASft-CR-177262) CODING FOB EfLIABLE N86-30035^SATELLITE COBHUNICailCRS final Report, 1'Jul. 1984 - 30 ODD.,1966 {Hawaii Univ.,iManoa.) 21 p CSCL 17B Unclas' G3/32 42931
https://ntrs.nasa.gov/search.jsp?R=19860020563 2018-04-16T20:03:22+00:00Z
CODING FOR RELIABLE SATELLITE COMMUNICATIONS
ABSTRACT
This research project was set up to investigate several error control
coding techniques for reliable satellite-communications. During the project
period/ we investigated the following areas: (1) decoding of- Reed^solomon
codes in terms of dual basis; (2) concatenated and cascaded error control
coding schemes for satellite and. space communications;—(3.).-using -hybrid, coding,
schemes (error correction and detection incorporated with retransmission) to
improve system reliability and throughput in satellite communications;
(4) good codes for simultaneous error correction and error detection; and
(5) error control techniques for ring and star networks. Significant results
were obtained in all the above areas.
I. A SUMMARY OP RESEARCH RESULTS
This research project was set up to study various kinds of coding
techniques for error control in satellite and space communications for NASA
Goddard Space Flight Center. During the project period, we investigated the
following areas: (1) decoding of Reed-Solomon codes in terms of dual basis;
(2) concatenated and cascaded error control .coding schemes for satellite and
space communications; (3) using hybrid coding schemes (error correction and
detection incorporated with retransmission) to improve system reliability and
throughput in satellite communications; (4) good codes for simultaneous error
correction and error detection; and (5) error control techniques for ring and
star networks. Significant results were obtained in all the above areas. In
the following, we summarize our research results.
1. Decoding of Reed-Solomon Codes in Dual Basis
Reed-Solomon codes form a class of very powerful cyclic block codes.
They are widely used for controlling transmission errors in data communication
systems as well as data storage systems. Recently Berlekamp [1] devised a new
method for encoding these codes which greatly reduces the encoding-complexity.
Berlekamp's encoder is implemented"in terms "of dual" basis using bit-serial
multipliers.
During the project period, we investigated decoding of Reed-Solomon codes
using the dual basis. The decoding-algorithm being used is-.the: Peterson-
Berklekamp-Chien algorithm. The algorithm consists of four steps.:
1. Compute the syndrome S = (S. ,S_,.. .S_) from-the.received ... •-- - -
polynomial r(X). - - - • -
2. Determine the error-location polynomial a(X) from the syndrome S.
3. Determine the error-value evaluator---Z(X) from the syndrome. -.-•.—-_.
4. Evaluate the error-location numbers and error values, and perform--
error correction. -
-2-
All the four decoding steps can be carried out in dual basis using bit-serial
multiplications or combination of bit-serial multiplications and parallel
multiplications. The circuit for the four decoding steps are- shown in Figures
1 and 3. An organization for a Reed-Solomon code decoder is shown in Figure 4.
A technical report on the decoding of Reed-Solomon codes in the dual form
was written and submitted to NASA Goddard Space Flight. Center.
2. A Concatenated Coding Scheme for NASA Telecommand System
During the project period, we also investigated a 'concatenated coding•----.
scheme for error control in data communications. In this scheme, the inner
code is used for both error correction and error detection, however the outer
code is used only for error detection. -A .retransmission-is requested if the
outer code detects the presence of errors after the inner code decoding —_.
Probability of undetected error is derived and bounded. A particular scheme
proposed for NASA Telecommand system is analyzed.
In the scheme proposed for NASA Telecommand system, both inner code and
outer code are shortened Hamming codes. The inner code is a distance-4
shortened Hamming code with generator polynomial,
g(X) = (X+1)(X6+X+1) = X7+X6+X2+1 .
This code is capable of correcting any single error and detecting any double
errors. The outer code is also a distance-4 shortened Hamming code with
generator polynomial,
g(X) = X16 +X12 + X5 + 1 .
This code is the X.25 standard for packet-switched data network-[2], The 16
parity bits of this code is used for error~detection-"only. The reliability
performance of the above scheme is_ -analyzed-*- We have-shown that, for —-—
bit-error-rate less than 10 , the" scheme provides extremely high.: --. • -
reliability.
-3-
A technical report on the performance study of the concatenated coding
scheme described above was written and sent to NASA Goddard Space Flight
Center.
3. A. Cascaded Error Control Coding Scheme for Satelliteand Space Communications
In this scheme, two linear block codes, C^ and C2, are used. The inner
code C^ is a binary (n,,k,) code~ with minimum. distance d.-. -The inner code is
designed to correct t^ or fewer-errors and simultaneously detect X,( Av > t ) --
or fewer errors where t,+X,+l <. d.. [3], The outer code C2 is an (n2,k2) code
£with symbols from the Galois Field GF(2 ) and minimum distance d_. If each
code symbol of the outer code is represented by binary £-tuple based on cer-
otain basis of GF(2 ), then the outer code becomes an (n2&,k2£) linear binary
code. For the proposed coding scheme, we assume -that the following conditions
hold:
and
where m and m are positive integers.
The encoding is performed in two stages as shown in Figure 5. First a
message of k2£ binary information digits is divided into -k^ bytes of
information bits each. Each 2,-bit byte (or binary Jl-tuple) is regarded as a
symbol in GF(2 ). These k_ bytes' are encoded according to the outer code
C to form an n_-byte (n_Jl bits) codeword in C . At ther second stage
of encoding, the n_-byte codeword at the output of the outer- code encoder is
divided into m. segments o_f m .bytes (or m^i bits)_each. Each m. byte seg-
ment is then encoded according to the^ inner code C. -to- form an n -bit code- •
word. This n -bit codeword in C _. is-called a frame. Thus^ corresponding to
message of k_£-bit at the input of- the- outer code encoder, -the output of the
inner code encoder is a sequence of nu frames of n. bits each. This ------
-4-
sequence of m2 frames is called a block. A block format is depicted in
Figure 6.
The decoding of the scheme also consists of two stages as shown in
Figure 5. The first stage of decoding is the inner- code decoding. Depending
on the number of errors in a received frame, the inner code decoder performs
one of the three following operations: error-correction, erasure^ and leave-
it-alone (LIA) operations. When a frame in _a block.-is~received, its syndrome
is computed based on the inner code-C-r—l-f-.the syndrome-corresponds to an
error pattern e of t. or fewer errors7'error correction is performed by
adding e to the received frame. The n - parity bits are removed from the
decoded frame, and the decoded m -byte segment-is stored in a receiver - --.-
buffer for the second stage of decoding. ..A successfully decoded segment is
called a decoded segment with no mark. Note-that the decoded segment is
error-free, if the number of transmission errors in a received frame is t ...
or less. If the number of transmission errors in a received frame is more
than \ , the errors may result in a syndrome which corresponds to a correct-
able error pattern with t. or fewer errors. In this case, the decoding will
be successful, but the decoded frame (or segment) contains .undetected errors.
If an uncorrectable error pattern is detected in a received frame, the inner
code decoder will perform one of the-following two operations based on a
certain criterion: - -- —
1. Erasure Operation — The erroneous-segment—rs~erased; We will call
such a segment an erased segment. - -- . -.„
2. Leave-it-alone (LIA) Operatic — The-erroneous-segment is-stored in"
the receiver buffer with a mark. -We-call-such a segment a marked
segment.
-5-
Thus, after m frames of a received block have been processed, the receiver
buffer may contain three types of segments: decoded segments without marks,
erroneous segments with marks, and erased segments.
As soon as m_ frames in a received block have been processed, the
second stage of decoding begins and the outer code decoder starts to decode
the m. segments stored in the buffer. -.Note-that an erased segment creates
m. symbol erasures (or m 5,-bit-byte erasures)-. Symbol erroxs;are -
contained in the segments with""or without marks. The outer code C_ and its
decoder are designed to correct the combinations"of symbol erasures and symbol
errors. Maximum-distance-separable codes with symbol from GF(2 ) are most
effective in correcting symbol erasures and errors.
Let i and h be the numbers of erased segments and marked segments
respectively. The outer code decoder declares an "erasure"(or raises'a flag)
for the entire block of m_ segments if either of the following two events
occurs:
(i) The number is is greater than a certain threshold T withes
T < (d_-l)/m..CS ^~ & X
(ii) The number h is greater than a certain-threshold T -.(-i)-with- - ------
Te£(i) <_ (d2-l-m1i)/2 for a given i.
If none of the above two events occurs', the outer code decoder starts the
error-correction operation on the m. decoded segments. The m..i symbol
erasures and the symbol errors in the marked or unmarked segments are
corrected based on the outer code C^i—Let t-2-(i-) be the error-correction
threshold for a given i where - _.i™r" ,.;_. .--..-. ... . .
If the syndrome of m_ decoded segments-in-the buffer corresponds to an error
pattern of m.i erasures and t_(i) or fewer symbol, errors, error-correction
is performed. The values of the-erased symbols, and the values'and~the
-6-
locations of symbol errors are determined based on a certain algorithm. If
more than t-(i) symbol errors are detected, then the outer code decoder
again declares an erasure (or raises a flag) for the entire block of m2
decoded segments.
The error performance of the proposed cascaded coding scheme where the -
outer code is used for both error correction'and'detection was analyzed. We
showed that, if proper inner and outer codes are chosen, the scheme provides .-
extremely good reliability even for high-bit-error-rate e=10 .- ~The scheme"
is particularly suitable for down.link error control in satellite communica-
tions. A number of specific schemes using various inner and outer codes,were
proposed to NASA-GSFC for possible applications in satellite communications.
A technical report on this scheme-was- submitted to NASA-GSFCr~
4. Error Detecting Capabilities of IEEE standard 802.3 Codes
During the project period, we investigated the-error detecting capabili-
ties of the shortened Hamming codes which are adopted for error detection in
IEEE Standard 802.3 CSMA/CD. These codes are also used for error detection in
the data link layer of the Ethernet,~a local—network. The generator polyno-
mial of these codes is X32+X2^+X23+X22+X16+X12+Xi:L+X10.4X8+X7--fX5+X4+X2+X+l, a -
primitive polynomial of degree 32. —Let C_-denote the shortened code of •--
length n. In the Ethernet the code length n is a multiple of 8 greater than
511 and less than 12145.
We first compute the weight distribution of-the dual-code~of~C""for
n=2p with 9<p<.13 and n=12144 by the Method -M-in- [41. -—osingFMacWilliams-1 -
identity we compute the number of codewords in C who'se~we~igfit:"Is" i..for -.... =
these n and 3<i<30. From the results we notice that the minimum-distance of —
-7-
CR, denoted d , is 4 or 5 for 512<n<12144. By finding the n such that d -5
and dn +1=4, we show that d^S for 512£n£3006, and d =4 for 3007£n<12144.
Let pe(cnȣ) and P(j(Cn,e) be the -probability of undetectable error and
that of detectable error, respectively, when code Cn-is used for error detec-
tion on a binary symmetric channel with bit-error-rate e. Prom the weight
distributions of the dual codes, we compute pe(cn»
£) and pd^cn'e' ^or n=2^
with 9<p£13 and n=12144, and 10~5-<e<l/-2. The results-are .plotted in. -Figures 7
and 8. From the computation we see that the maximum value of P-e 5 1'2 »"£"")
e<l/2 is 2.6544E-10 which occurs for e=1.3918E-2 and -the maximum value of P '"
Pe(C1024,e) for e<l/2 is 2.3286E-10 which occurs for £>=1.4383E-2. Note that
these peak values are greater than 2 . For "the larger values n=2^ with
115E5.13 and n=12444, no peak is detected iwithin accuracy in computation.
We also analyze the double-burst-detecting capability. Using the algo-
rithm in [5] we compute the maximum code length n_ such that Cn has the
capability of correcting any burst error of length b or less. We show that
n =38 for 14<b<16, n, =730, n =1729, n =5680, n =11933 and n <13000.D -- i j 12 . 11 - - - 10 gr~
A technical report is in preparation- and will-be -submitted -to "NASA-GSFG.-
5. An ARQ Scheme for Broadcast Communication Systems --_ _ _ .. -
Consider a point-to-multipoint communication system consisting of (R+l)
stations, where a single transmitter broadcasts data frames -to R receivers,
each of which has a finite buffer capacity to store data frames for
processing. During the project period,- we investigated -an- ARQ scheme "for ---—
error control in such system. In our proposed schemeyeach-data frame ----- =--r -
consists of k message bits and .(n-rk)-~parity bits-j*hieh_are for¥ed~>ase~cLjon:
an (n,k) linear block code for errot, detection. -When a data frame is received
by a receiver, parity checking is. .per formed:.".: If no error is -detected, the
received data frame (with n-k parity -bits removed)-iS:~either -delivered to the
user or stored in the receiver buf-fer until-it is ready to be delivered to the
-8-
user. If a received frame is detected in errors, it is discarded and the
receiver requests a retransmission of that frame. In our proposed retransmis-
sion strategy, we use a constraint in the transmitter :to prevent any buffer
overflow at the receivers. Retransmissions-continue until'positive acknow-
ledgements are received from all R receivers.- All the receivers that have
received a frame successfully, continue to positively acknowledge the retrans-
missions, whether or not the new copies of data frame are error free. Hence
the scheme makes full use of the outcomes of previous transmissions. The
proposed scheme can also handle data and/or acknowledgement loss. -
The throughput performance of the proposed scheme is analyzed and
simulated. Results obtained from analysis and simulation agree reasonably
well. The results alo show .that the proposed scheme outperforms .the full-
memory go-back-N scheme proposed by Gopal, _et al. {6],
A technical report on this ARQ scheme has been submitted to NASA-GSFC.
REFERENCES
1. E.R. Berlekamp, "Bit-Serial Reed-Solomon Encoders," IEEE Transactions onInformation Theory, Vol. IT-28, pp. 869-874, November,- .982. — —
2. CCITT: Recommendation X.25,"Interface-Between Data Terminal Equipmentand Data Circuit-Terminating Equipment.for..Terminals Operating in PacketMode on Public Data Networks," with Plenary Assembly, Doc. No. 7, Geneva,1980.
3. s. Lin and D.J. Costello, Jr., Error Control Coding: Fundamentals andApplications, Prentice-Hall, 1982- „. .- —-.--•- - --• —
4. T. Fujiwara, T. Kasami, A. Kitai and .S. :Lin, ~"0n the. Undetected Error -Probability for Shortened Hamming-Codes;-" IEEE -Transactions-on-— " - ~~-Communications, Vol. COM-33, No_._ _pp.--5.7:Q-&74-r(:June--i985)..•'-_- ._.-.- —i:
5. T. Kasami, "Optimum Shortened Cyclic -Codes for Correction,"—IEEE/- - —--•-*---Transactions on Information Theory,,..Vol. IT-9, No. 2,-pp.—105--109(April 1963).
6. Inder, S. Gopal and Jeffrey M. JaLfe,.•Point-to^Multipoint^Communication-over Broadcast Links," IEEE Trans. Comm.;. VolV"COM-J2tJ-No. 9, .Sept. -1-984.
-9-
II. PERSONNEL
Principal Investigators;
Dr. Shu Lin (July 1, 1984 - December 31, 1985; Dr. Lin is currently onleave)
Dr. N.T. Gaarder (January 1 - June 30, 1986)
Consultant; Professor Tadao Kasami
Graduate Students;
Ram ChandranMao-chao Lin - -
III. RESEARCH ACTIVITIES ".:""••- '--
Journal Publications . _ ... ' " • . - • .
1. "On the Probability of Undetected Error for the Maximum DistanceSeparable Codes," IEEE Transactions on-Communications^ Vol. COM-32^,No. 9, September, 1984.
2. "On the Undetected Error Probability for Shortened Hamming Codes,' IEEETransactions on Communications, Vol. COM-33, No. 6, June 1985.
3. 'A Concatenated Coding Scheme for Error Control," IEEE Transactions onCommunications, Vol. COM-34, No. 6, May 1986.
4. "A Cascaded Coding Scheme for Error Control," to appear in IEEETransactions on Information Theory, 1987.
5. "Error Detecting Capabilities of Shortened Hamming Codes Adopted forError Detection in IEEE Standard 802.3," Journal of "the Institute of ~Electronics and Communication Engineers of Japan, 4986.—- _=—--.
6. "On the Weight Distribution of Some Binary Reed-Solomon Codes," submittedto IEEE Transactions on Communications, (in revision) 1986.
Technical Reports
1. "Encoding and Decoding of Reed-Solomon. Codes in-Dual.Basis," TechnicalReport, NASA Grant NAG 5-407, October, 1984.- " y :::.'
2. "Probability of Undetected Error After Decoding..for.a-Concatenated CodingScheme for Error Control," Technical-Report NASA~Grant";5H407,;;July7~ 198*.,
3, "A Cascaded Coding Scheme for Error- Control," Technical—Report-I-NASAGrant 5-407 si, October 1985. =-- — - •--- - : , - - - -:-
4, "A Cascaded Coding Scheme for-Efror Control,7" Technical- ReportrII~NASAr rrGrant NAG 50496 SI, December 1985v - "~ . _ . _ - _ . .
-10-
5. 'A Selective Repeat ARQ Scheme for Point-to-Multipoint Communications andIts Throughput Analysis,* Technical Report, NASA Grant MAG 5-407 SI,June 13, 1986.
Conference Presentations
1. 'Encoding and Decoding of Reed-Solomon Codes in Dual Basis," Seminar,Osaka University, October 31, 1984.
2. "Probability of Undetected Error after Decoding for a Concatenated CodingScheme," 7th Conference on Information Theory and Its Applications,Kinaugawa, Japan, November 4-7, 1984.
3. "Future Problems in Coding"__(guest speaker), 7th Conference onInformation Theory and Its Applications,'Kinaugawa, Japan, November 4-7,1984.
4. "Performance Analysis of a Concatenated Coding Scheme for Error Control,"1984 Globecom, Atlanta, Georgia, November 26-29, 1984.
5. "On the Weight Distribution of Some Binary Reed-Solomon Codes,"Proceedings of IEEE International Symposium on Information--Theory_(Abstract), Brighton, England, June, 1985. ----
6. "A Concatenated Coding Scheme for Error Control," Proceedings of IEEEInternational Symposium on Information Theory (Abstract), Brighton,England, June, 1985.
7. "A Survey of Error Control Schemes for Space and Satellite Communica-tions" (Invited paper) The 8th Symposium on Information Theory and ItsApplications, Nara, Japan, December 5-7, 1985.
8. "The Probabilities of Correct Decoding^ Decoding Error and DecodingFailure for a Cascaded Coding Scheme,* Proceedings of the 8th Symposiumon Information Theory and Its Applications, Nara, Japan, December 5-7,—1985.
9. "Two Classes of Codes for Unequal Error Protection," to be presented atthe IEEE International Symposium on Information-Theory, -Ann Arbor, -Michigan, October 5-9, 1986* — :-••
10. "A Cascaded Coding Scheme for Error Control," to be presented at the IEEEInternational Symposium on Informat-ion Theory>- Ann Arbor,-Michigan, -- - - —October 5-9, 1986.
IT. --"Error Detecting Capabilities of-the: Shortened-Hamming Codes-Adopted.-for _Error Detection in IEEE standard-802T3,"-<L.tO-be-presented~at -the IEEEInternational Symposium on Information -Thfr&r-y .Ann.-Ai:b&r-f--Michigan-, --...;r ,_:October 5-9, 1986.
12. "A Selective Repeat ARQ Scheme for.Point-to-Multipoint Communications andIts Throughput Analysis," ACM Conference-on--Gommunicati'on-Protocols,Vermont, August 1986.
-11-
13. 'A Cascaded Error Control Coding Scheme for Space and SatelliteCommunications," submitted to Globecom-1986, Houston, Texas, December,1986.
Consultation with NASA Officers
During the project period, Dr. Lin made'three visits (November 3, 1984;
March 29, 1985 and April 1, 1986) to NASA Goddard Space Plight Center,-and
discussed with Dr. James C. Norakis and Mr. Warner H. Miller on-various
project problems.
-12-
r(*}TVA )
•
•
1
N
... 1
*Bit-SerialMultiplier
ultiply by
~^/1
\
a
\
\(\^
r
^J
f
[Lr,
-
\ »^
-
i
1
-..
|
[
-
- - - -
-_ „.. .
-K4JFigure 1 A circuit for computing -the -syndrome-.component
S. with a bit-serial multiplier- -
*-R»gt>ter (dual for») •-Register (dual fora)
Control Circuit, o(u). e(v) <
i
font)
i : Syndrow Realster:(convention*) for*)
iFigure 2 Circuit for finding/o(X) and 2(X) O O•n 30
i i
O >
O '
i3
InputReceiver Buffer Register r(X) (dual form)
7.1£
n
*
2 2i201\
S
!
^
Z3Q3£\
fc \
i
T O • *t-Register(dual form)
• • •i I
1
AdderC
tin -
2(a£)
e(X)(dual form)
(dualhr
(conventional)
Gate
T2
f3 /conventional\« \ form j
- - r —* i / * \ T " 1[a o'(a )J/conventional\\ form /
(dual)Figure 3 Error-correction circuit-
r(X)
(X) Register
(X) Register
Buffer Register f(X)
SyndromeComputation
Galois Field
Arithmetic
Processor
o(X)
2(X)
Chien SearchCircuit i
Error-Magnitude
Computation
Output
V r{X)-e(X)'
e(X)
Gate
Figure 4 An organization of a RS decoder
Outer CodeEncoder
Inner Code.....Encoder
Channel
Outer CodeDecoder
1
Inner CodeDecoder
Figure 5 A cascaded coding system
I I
Frames
m -symbol Segment
L>£ Outer Code Parity Symbols
Frame
"Inner CodeParity Bits
Figure 6" Block format
The probobilty ofundetectable error
Pe(Cn
10-9 .
10-10-
10-11-1
10-12-
10-13-
10-14-
10-15
r5
n= 512n= 1024n= 2048n= 4096n= 8192n=12144
10-3 10-2
Bit-error-rate
10"1 1/2
Figure 7 The probability.Pe(Cn,tT~£KSta lrecTS±ved-vex;tor"contains an undetected error pattern for—a -—binary symmetric channel with bit-er-ror^rate e.~
The probability ofdetectable error
pd(cn 'e)
1-
10"1-
HT*-
10-3
n= 512n= 1024n= 2048n= 4096n= 8192n=12144
10' 10,-4 10,-3 1/2
Bit-error-rate
Figure 8 The probability Pd(Cn,e)= that a"-reee'ivea-vectorcontains a detectable error pattern" "for a binarysymmetric channel with bit-error-rate~£. - -"—-"—